DOCSLIB.ORG

Information Measures for Deterministic Input-Output Systems

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Information Measures for Deterministic Input-Output Systems 1 Information Measures for Deterministic Input-Output Systems Bernhard C. Geiger, Student Member, IEEE, and Gernot Kubin, Member, IEEE Abstract—In this work the information loss in deterministic, processing, just as, loosely speaking, a passive system cannot memoryless systems is investigated by evaluating the conditional increase the energy contained in a signal1. entropy of the input random variable given the output random Clearly, the information lost in a system not only depends variable. It is shown that for a large class of systems the infor- mation loss is finite, even if the input is continuously distributed. on the system but also on the signal carrying this information; Based on this finiteness, the problem of perfectly reconstructing the same holds for the energy lost or gained. While there the input is addressed and Fano-type bounds between the is a strong connection between energy and information for information loss and the reconstruction error probability are Gaussian signals (entropy and entropy rate of a Gaussian derived. signal are related to its variance and power spectral density, For systems with infinite information loss a relative measure is defined and shown to be tightly related to Rényi information respectively), for non-Gaussian signals this connection degen- dimension. Employing another Fano-type argument, the recon- erates to a bound by the max-entropy property of the Gaussian struction error probability is bounded by the relative information distribution. Energy and information of a signal therefore can loss from below. behave completely differently when fed through a system. But In view of developing a system theory from an information- while we have a definition of the energy loss (namely, the theoretic point-of-view, the theoretical results are illustrated by inverse gain), an analysis of the information loss in a a few example systems, among them a multi-channel autocorre- L2 lation receiver. system is still lacking. It is the purpose of this work to close this gap and to Index Terms—Data processing inequality, Fano’s inequality, information loss, Rényi information dimension, system theory propose the information loss – the conditional entropy of the input given the output – as a general system characteristic, complementing the prevailing energy-centered descriptions. I. INTRODUCTION The choice of this conditional entropy is partly motivated by When opening a textbook on linear [1] or nonlinear [2] the data processing inequality (cf. Definition 1) and justified input-output systems, the characterizations one typically finds by a recent axiomatization of information loss [6]. – aside from the difference or differential equation defining At present we restrict ourselves to memoryless systems2 the system – are almost exclusively energy-centered in nature: operating on (multidimensional) random variables. The reason transfer functions, input-output stability, passivity, lossless- for this is that already for this comparably simple system class ness, and the or energy/power gain are all defined using the L2 a multitude of questions can be asked (e.g., what happens if we amplitudes (or amplitude functions) of the involved signals, lose an infinite amount of information?), to some of which we therefore essentially energetic in nature. When opening a intend to present adequate answers (e.g., by the introduction textbook on statistical signal processing [3] or an engineering- of a relative measure of information loss). This manuscript can oriented textbook on stochastic processes [4], one can add thus be regarded as a first small step towards a system theory correlations, power spectral densities, and how they are af- from an information-theoretic point-of-view; we hope that the fected by linear and nonlinear systems (e.g., the Bussgang results presented here will ley the foundation for some of the arXiv:1303.6409v2 [cs.IT] 17 Apr 2013 theorem [4, Thm. 9-17]). By this overwhelming prevalence of forthcoming steps, such as extensions to stochastic processes energetic measures and second-order statistics, it is no surprise or systems with memory. that many problems in system theory or signal processing are formulated in terms of energetic cost functions, e.g., the mean- squared error. A. Related Work What one does not find in all these books is an information- To the best of the authors’ knowledge, very few results theoretic characterization of the system at hand, despite the about the information processing behavior of deterministic fact that such a characterization is strongly suggested by an input-output systems have been published. Notable exceptions elementary theorem: the data processing inequality. We know are Pippenger’s analysis of the information lost in the mul- that the information content of a signal (be it a random vari- tiplication of two integer random variables [7] and the work able or a stochastic process) cannot increase by deterministic of Watanabe and Abraham concerning the rate of informa- tion loss caused by feeding a discrete-time, finite-alphabet Bernhard C. Geiger ([email protected]) and Gernot Kubin ([email protected]) are with the Signal Processing and Speech stationary stochastic process through a static, non-injective Communication Laboratory, Graz University of Technology Parts of this work have been presented at the 2011 IEEE Int. Sym. 1At least by not more than a finite amount, cf. [5]. on Wireless Communication Systems, the 2012 Int. Zürich Seminar on 2Since these systems will be described by (typically) nonlinear functions, Communications, and the 2012 IEEE Information Theory Workshop. we will use “systems” and “functions” interchangeably. 2 function [8]. Moreover, in [9] the authors made an effort to B. Outline and Contributions extend the results of Watanabe and Abraham to dynamical The organization of the paper is as follows: In Section II we input-output systems with finite internal memory. All these give the definitions of absolute and relative information loss works, however, focus only on discrete random variables and and analyze their elementary properties. Among other things, stochastic processes. we prove a connection between relative information loss and Slightly larger, but still focused on discrete random vari- Rényi information dimension and the additivity of absolute ables only, is the field concerning information-theoretic cost information loss for a cascade. We next turn to a class of functions: The infomax principle [10], the information bottle- systems which has finite absolute information loss for a real- neck method [11] using the Kullback-Leibler divergence as valued input in Section III. A connection to differential entropy a distortion function, and system design by minimizing the is shown as well as numerous upper bounds on the information error entropy (e.g., [12]) are just a few examples of this recent loss. Given this finite information loss, we present Fano- trend. Additionally, Lev’s approach to aggregating accounting type inequalities for the probability of a reconstruction error. data [13], and, although not immediately evident, the work Section IV deals with systems exhibiting infinite absolute about macroscopic descriptions of multi-agent systems [14] information loss, e.g., quantizers and systems reducing the belong to that category. dimensionality of the data, to which we apply our notion A system theory for neural information processing has been of relative information loss. Presenting a similar connection proposed by Johnson3 in [15]. The assumptions made there between relative information loss and the reconstruction er- (information need not be stochastic, the same information can ror probability establishes a link to analog compression as be represented by different signals, information can be seen investigated in [26]. We apply our theoretical results to two as a parameter of a probability distribution, etc.) suggest the larger systems in Section V, a communications receiver and use of the Kullback-Leibler divergence as a central quantity. an accumulator. It is shown that indeed both the absolute and Although these assumptions are incompatible with ours, some relative measures of information loss are necessary to fully similarities exist (e.g., the information transfer ratio of a characterize even the restricted class of systems we analyze in cascade in [15] and Proposition 4). this work. Eventually, Section VI is devoted to point at open issues and lay out a roadmap for further research. Aside from these, the closest the literature comes to these concepts is in the field of autonomous dynamical systems or iterated maps: There, a multitude of information-theoretic II. DEFINITION AND ELEMENTARY PROPERTIES OF characterizations are used to measure the information transfer INFORMATION LOSS AND RELATIVE INFORMATION LOSS in deterministic systems. In particular, the information flow A. Notation between small and large scales, caused by the folding and We adopt the following notation: Random variables (RVs) stretching behavior of chaotic one-dimensional maps was are represented by upper case letters (e.g., ), lower case analyzed in [16]; the result they present is remarkably sim- X letters (e.g., ) are reserved for (deterministic) constants or ilar to our Proposition 7. Another notable example is the x realizations of RVs. The alphabet of an RV is indicated by a introduction of transfer entropy in [17], [18] to capture the calligraphic letter (e.g., ). The probability distribution of an information transfer between states of different (sub-)systems. RV is
Load more

Recommended publications
  • Polarization of the Rényi Information Dimension with Applications To
    1 Polarization of the Renyi´ Information Dimension with Applications to Compressed Sensing Saeid Haghighatshoar, Member, IEEE, Emmanuel Abbe, Member, IEEE Abstract—In this paper, we show that the Hadamard matrix in diverse areas in probability theory and signal processing acts as an extractor over the reals of the Renyi´ information such as signal quantization [4], rate-distortion theory [5], and dimension (RID), in an analogous way to how it acts as fractal geometry [6]. More recently, the operational aspect of an extractor of the discrete entropy over finite fields. More precisely, we prove that the RID of an i.i.d. sequence of mixture RID has reappeared in applications as varied as lossless analog random variables polarizes to the extremal values of 0 and 1 compression [7–9], Compressed Sensing of sparse signals [10– (corresponding to discrete and continuous distributions) when 12], and characterization of the degrees-of-freedom (DoFs) of transformed by a Hadamard matrix. Further, we prove that vector interference channels [13, 14], which has recently been the polarization pattern of the RID admits a closed form of significant importance in wireless communication. expression and follows exactly the Binary Erasure Channel (BEC) polarization pattern in the discrete setting. We also extend the In this paper, motivated by [3], we first extend the definition results from the single- to the multi-terminal setting, obtaining a of RID as an information measure from scalar random Slepian-Wolf counterpart of the RID polarization. We discuss variables to a family of vector random variables over which applications of the RID polarization to Compressed Sensing the RID is well-defined.
    [Show full text]
  • Rényi Information Dimension: Fundamental Limits of Almost Lossless Analog Compression Yihong Wu, Student Member, IEEE, and Sergio Verdú, Fellow, IEEE
    IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 56, NO. 8, AUGUST 2010 3721 Rényi Information Dimension: Fundamental Limits of Almost Lossless Analog Compression Yihong Wu, Student Member, IEEE, and Sergio Verdú, Fellow, IEEE Abstract—In Shannon theory, lossless source coding deals with • The central problem is to determine how many compressed the optimal compression of discrete sources. Compressed sensing measurements are sufficient/necessary for recovery with is a lossless coding strategy for analog sources by means of mul- vanishing block error probability as blocklength tends to tiplication by real-valued matrices. In this paper we study almost lossless analog compression for analog memoryless sources in an infinity [2]–[4]. information-theoretic framework, in which the compressor or de- • Random coding is employed to show the existence of compressor is constrained by various regularity conditions, in par- “good” linear encoders. In particular, when the random ticular linearity of the compressor and Lipschitz continuity of the projection matrices follow certain distribution (e.g., stan- decompressor. The fundamental limit is shown to be the informa- dard Gaussian), the restricted isometry property (RIP) is tion dimension proposed by Rényi in 1959. satisfied with overwhelming probability and guarantees Index Terms—Analog compression, compressed sensing, infor- exact recovery. mation measures, Rényi information dimension, Shannon theory, On the other hand, there are also significantly different ingre- source coding. dients in compressed sensing in comparison with information theoretic setups. I. INTRODUCTION • Sources are not modeled probabilistically, and the funda- mental limits are on a worst case basis rather than on av- A. Motivations From Compressed Sensing erage. Moreover, block error probability is with respect to the distribution of the encoding random matrices.
    [Show full text]
  • Understanding the Fractal Dimensions of Urban Forms Through Spatial Entropy
    entropy Article Understanding the Fractal Dimensions of Urban Forms through Spatial Entropy Yanguang Chen 1,* ID , Jiejing Wang 2 and Jian Feng 1 1 Department of Geography, College of Urban and Environmental Sciences, Peking University, Beijing 100871, China; [email protected] 2 Department of Urban Planning and Management, School of Public Administration and Policy, Renmin University of China, Beijing 100872, China; [email protected] * Correspondence: [email protected]; Tel.: +86-106-276-6277 Received: 17 September 2017; Accepted: 6 November 2017; Published: 9 November 2017 Abstract: The spatial patterns and processes of cities can be described with various entropy functions. However, spatial entropy always depends on the scale of measurement, and it is difficult to find a characteristic value for it. In contrast, fractal parameters can be employed to characterize scale-free phenomena and reflect the local features of random multi-scaling structure. This paper is devoted to exploring the similarities and differences between spatial entropy and fractal dimension in urban description. Drawing an analogy between cities and growing fractals, we illustrate the definitions of fractal dimension based on different entropy concepts. Three representative fractal dimensions in the multifractal dimension set, capacity dimension, information dimension, and correlation dimension, are utilized to make empirical analyses of the urban form of two Chinese cities, Beijing and Hangzhou. The results show that the entropy values vary with the measurement scale, but the fractal dimension value is stable is method and study area are fixed; if the linear size of boxes is small enough (e.g., <1/25), the linear correlation between entropy and fractal dimension is significant (based on the confidence level of 99%).
    [Show full text]
  • Information Theory and Dynamical System Predictability
    Entropy 2011, 13, 612-649; doi:10.3390/e13030612 OPEN ACCESS entropy ISSN 1099-4300 www.mdpi.com/journal/entropy Article Information Theory and Dynamical System Predictability Richard Kleeman Courant Institute of Mathematical Sciences, New York University, 251 Mercer Street, New York, NY 10012, USA; E-Mail: [email protected] Received: 25 January 2011; in revised form: 14 February 2011 / Accepted: 20 February 2011 / Published: 7 March 2011 Abstract: Predicting the future state of a turbulent dynamical system such as the atmosphere has been recognized for several decades to be an essentially statistical undertaking. Uncertainties from a variety of sources are magnified by dynamical mechanisms and given sufficient time, compromise any prediction. In the last decade or so this process of uncertainty evolution has been studied using a variety of tools from information theory. These provide both a conceptually general view of the problem as well as a way of probing its non-linearity. Here we review these advances from both a theoretical and practical perspective. Connections with other theoretical areas such as statistical mechanics are emphasized. The importance of obtaining practical results for prediction also guides the development presented. Keywords: predictability; information theory; statistical physics 1. Introduction Prediction within dynamical systems originated within the modern era in the study of the solar system. The regularity of such a system on time scales of centuries meant that very precise predictions of phenomena such as eclipses are possible at such lead times. On longer times scales of order million or more years, chaotic behavior due to the multi-body gravitational interaction becomes evident (see e.g., [1]) and effects such as planetary collisions or system ejection can occur.
    [Show full text]
  • Understanding Fractal Dimension of Urban Form Through Spatial Entropy
    Understanding Fractal Dimension of Urban Form through Spatial Entropy Yanguang Chen1; Jiejing Wang2; Jian Feng1 (1. Department of Geography, College of Urban and Environmental Sciences, Peking University, Beijing 100871, P.R. China. 2. Department of Urban Planning and Management, School of Public Administration and Policy, Renmin University of China, Beijing, P.R. China. E-mail: [email protected]; [email protected]; [email protected]) Abstract: Spatial patterns and processes of cities can be described with various entropy functions. However, spatial entropy always depends on the scale of measurement, and it is difficult to find a characteristic value for it. In contrast, fractal parameters can be employed to characterize scale-free phenomena. This paper is devoted to exploring the similarities and differences between spatial entropy and fractal dimension in urban description. Drawing an analogy between cities and growing fractals, we illustrate the definitions of fractal dimension based on different entropy concepts. Three representative fractal dimensions in the multifractal dimension set are utilized to make empirical analyses of urban form of two cities. The results show that the entropy values are not determinate, but the fractal dimension value is certain; if the linear size of boxes is small enough (e.g., <1/25), the linear correlation between entropy and fractal dimension is clear. Further empirical analysis indicates that fractal dimension is close to the characteristic values of spatial entropy. This suggests that the physical meaning of fractal dimension can be interpreted by the ideas from entropy and scales and the conclusion is revealing for future spatial analysis of cities. Key words: fractal dimension; entropy; mutlifractals; scaling; urban form; Chinese cities 1.
    [Show full text]
  • Information, Measure Shifts and Distribution Diagnostics
    May 2007 Information, measure shifts and distribution diagnostics Roger J. Bowden* Summary Reflexive shifting of a given distribution, using its own distribution function, can reveal information. The shifts are changes in measure accomplished via simple Radon-Nikodym derivatives. The separation of the resulting left and right unit shifted distributions reveals the binary entropy of location. Recombining the shifts via convex combinations yields information-based asymmetry and spread metrics as well as insight into latent structure. The resulting diagnostics are applicable even for long tail densities where distributional moments may not exist, and independently also of benchmarks such as the Gaussian. Key Words: Asymmetry, entropy bandwidth, locational entropy, distribution spread, reflexive shifting, unit distribution shifts. MSC Classifications: 60E05; 94A17. *School of Economics and Finance, Victoria University of Wellington, and Kiwicap Research Ltd. Emails [email protected]; [email protected] . Phone + 64 4 472 4984, fax 4983, postal address School of Economics and Finance, Victoria University of Wellington, P.O. Box 600, Wellington, New Zealand. 1. Introduction The central idea developed in this paper is that shifting a given distribution in a reflexive manner, by using its own distribution function, can reveal information. The resulting operation can be applied in such contexts as latent mixtures to reveal more of the underlying structure. The shift framework can also be useful in originating metrics relating to spread and asymmetry of a distribution. The latter generally assume a benchmark such as the normal distribution, or that symmetry is just a matter of odd-order moments. But in very long tail densities such as the Levy distribution, excess kurtosis and asymmetry metrics based on moments do not even exist.
    [Show full text]
  • Measuring Statistical Dependence Via the Mutual Information Dimension
    Proceedings of the Twenty-Third International Joint Conference on Artificial Intelligence Measuring Statistical Dependence via the Mutual Information Dimension Mahito Sugiyama1 and Karsten M. Borgwardt1;2 1Machine Learning and Computational Biology Research Group, Max Planck Institute for Intelligent Systems and Max Planck Institute for Developmental Biology, Tubingen,¨ Germany 2Zentrum fur¨ Bioinformatik, Eberhard Karls Universitat¨ Tubingen,¨ Germany fmahito.sugiyama, [email protected] Abstract al. [2011] (further analyzed in [Reshef et al., 2013]) that mea- sures any kind of relationships between two continuous vari- We propose to measure statistical dependence be- ables. They use the mutual information obtained by discretiz- tween two random variables by the mutual in- ing data and, intuitively, MIC is the maximum mutual infor- formation dimension (MID), and present a scal- mation across a set of discretization levels. able parameter-free estimation method for this task. However, it has some significant drawbacks: First, MIC Supported by sound dimension theory, our method depends on the input parameter B(n), which is a natural gives an effective solution to the problem of detect- number specifying the maximum size of a grid used for dis- ing interesting relationships of variables in massive cretization of data to obtain the entropy [Reshef et al., 2011, data, which is nowadays a heavily studied topic in SOM 2.2.1]. This means that MIC becomes too small if we many scientific disciplines. Different from classi- choose small B(n) and too large if we choose large B(n). cal Pearson’s correlation coefficient, MID is zero Second, it has high computational cost, as it is exponential if and only if two random variables are statistically with respect to the number of data points1, and not suitable independent and is translation and scaling invari- for large datasets.
    [Show full text]
  • Information Dimension and the Probabilistic Structure of Chaos J
    Information Dimension and the Probabilistic Structure of Chaos J. Doyne Farmer Dynamical Systems Collective, Physics Department, UC Santa Cruz, Santa Cruz, California Z. Naturforsch. 37a, 1304-1325 (1982); received May 18, 1982 The concepts of entropy and dimension as applied to dynamical systems are reviewed from a physical point of view. The information dimension, which measures the rate at which the in- formation contained in a probability density scales with resolution, fills a logical gap in the classification of attractors in terms of metric entropy, fractal dimension, and topological entropy. Several examples are presented of chaotic attractors that have a selfsimilar, geometrically scaling structure in their probability distribution; for these attractors the information dimension and fractal dimension are different. Just as the metric (Kolmogorov-Sinai) entropy places an upper bound on the information gained in a sequence of measurements, the information dimension can be used to estimate the information obtained in an isolated measurement. The metric entropy can be expressed in terms of the information dimension of a probability distribution constructed from a sequence of measurements. An algorithm is presented that allows the experimental de- termination of the information dimension and metric entropy. Introduction the natural vehicle to make the notion of "degrees of freedom" more precise. Let us start out with a quotation from Abbot's Following A Square [1], we will distinguish three "Fiatland" [1]: distinct intuitive notions of dimension: "Dimension implies direction, implies measurement, implies the more and the less." (1) Direction: Euclidean 3-space is naturally de- scribed by 3-tuples, and is not homeomorphic to "As one of your Spaceland poets has said, we are all liable to the same errors, all alike the Slaves of our re- Euclidean 2-space.
    [Show full text]
  • Modeling and Analyzing Fractal Point Processes by Anand R
    RICE UNIVERSITY Modeling and Analyzing Fractal Point Processes by Anand R. Kumar A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy Approved, Thesis Committee: Professor of Electrical and Computer Engineering John \V. Clark Professor of Electrical and Computer Engineering r· . i ' .• (// -; ,,<(, T ·.· (I {· ! .••• : /j _/ Katherine B. Ensor Assistant Professor of Statistics Chiyeko Tsuchitani Adjunct Professor of Electrical and Computer Engineeering Houston, Texas April, 1990 3 1272 00637 7038 Modeling and Analyzing Fractal Point Processes Anand R. Kumar Abstract The temporal pattern of the action potential discharges of auditory nerve fibers have been presumed to be random in nature and well described as a renewal process. Recently, we observed that the discharges were characterized by excess variability and low frequency spectra. Simple stationary point process models, point process models with a chaotic intensity, and doubly stochastic point process models with a strongly mixing intensity process do not describe the data. But a fractal point process, defined to be a doubly stochastic point process with a fractal waveform intensity process, described the the data more generally. The sample paths of the counting process of a fractal point process are not self-similar; it displays self-similar characteristics only over large time scales. The Fano factor, rather than the power spectrum, captures these self-similar char­ acteristics. The fractal dimension and fractal time characterize the Fano factor. The fractal dimension is characteristic of the fractal intensity process and is measured from the asymptotic slope of the Fano factor plot. The fractal time is the time before the onset of the fractal behavior and delineates the short-term characteristics of the 11 data.
    [Show full text]
  • Information-Theoretic Analysis of Memoryless Deterministic Systems
    Article Information-Theoretic Analysis of Memoryless Deterministic Systems Bernhard C. Geiger 1,* and Gernot Kubin 2 1 Institute for Communications Engineering, Technical University of Munich, Munich 80290, Germany 2 Signal Processing and Speech Communication Laboratory, Graz University of Technology, Graz 8010, Austria; [email protected] * Correspondence: [email protected]; Tel.: +49-89-289-23452 Academic Editors: Raúl Alcaraz Martínez and Kevin H. Knuth Received: 18 August 2016; Accepted: 14 November 2016; Published: 17 November 2016 Abstract: The information loss in deterministic, memoryless systems is investigated by evaluating the conditional entropy of the input random variable given the output random variable. It is shown that for a large class of systems the information loss is finite, even if the input has a continuous distribution. For systems with infinite information loss, a relative measure is defined and shown to be related to Rényi information dimension. As deterministic signal processing can only destroy information, it is important to know how this information loss affects the solution of inverse problems. Hence, we connect the probability of perfectly reconstructing the input to the information lost in the system via Fano-type bounds. The theoretical results are illustrated by example systems commonly used in discrete-time, nonlinear signal processing and communications. Keywords: information processing; signal processing; system theory; Fano’s inequality; information loss; Rényi information dimension 1. Introduction When opening a textbook on linear [1] or nonlinear [2] deterministic signal processing, input-output systems are typically characterized—aside from the difference or differential equation defining the system—by energy- or power-related concepts: L2 or energy/power gain, passivity, losslessness, input-output stability, and transfer functions are all defined using the amplitudes (or amplitude functions like the squared amplitude) of the involved signals, therefore essentially energetic in nature.
    [Show full text]